Journal of Experimental Psychology:
Human Perception and Performance
2010, Vol. 36, No. 3, 673– 688
© 2010 American Psychological Association
0096-1523/10/$12.00 DOI: 10.1037/a0016888
Crowding Affects Letters and Symbols Differently
Jonathan Grainger
Ilse Tydgat and Joanna Isselé
CNRS and Aix-Marseille University
Ghent University
Five experiments examined crowding effects with letter and symbol stimuli. Experiments 1 through 3
compared 2-alternative forced-choice (2AFC) identification accuracy for isolated targets presented left
and right of fixation with targets flanked either by 2 other items of the same category or a single item
situated to the right or left of targets. Interference from flankers (crowding) was significantly stronger for
symbols than letters. Single flankers generated performance similar to the isolated targets when the
stimuli were letters but closer to the 2-flanker condition when the stimuli were symbols. Experiment 4
confirmed this pattern using a partial-report bar probe procedure. Experiment 5 showed that another
measure of crowding, critical spacing, was greater for symbols than for letters. The results support the
hypothesis that letter-string processing involves a specialized system developed to limit the spatial extent
of crowding for letters in words.
Keywords: crowding, letter perception, orthographic processing
a function of eccentricity (Hammond & Green, 1982; Mason,
1982; Mason & Katz, 1978). Recent research has shown that this
phenomenon also holds for data-limited identification paradigms
(Tydgat & Grainger, 2009), suggesting that it is related to fundamental differences in the way strings of letters and symbols are
processed. The present study is a further attempt at understanding
the mechanisms that underlie such differences.
The standard serial position function for letter stimuli (typically
obtained by using random strings of consonants) is M-shaped for
response times in the letter search task (Hammond & Green, 1982;
Ktori & Pitchford, 2008, 2009; Mason, 1975, 1982; Mason &
Katz, 1976; Pitchford, Ledgeway, & Masterson, 2008) and
W-shaped for studies using data-limited identification procedures
(Averbach & Coriell, 1961; Butler, 1975; Butler & Merikle, 1973;
Haber & Standing, 1969; Merikle, Coltheart, & Lowe, 1971;
Merikle, Lowe, & Coltheart, 1971; Mewhort & Campbell, 1978;
Schwantes, 1978; Stevens & Grainger, 2003; Wolford & Hollingsworth, 1974). One classic interpretation of this serial position
function is that it reflects the combination of two factors: the drop
of acuity from fixation to the periphery, and less crowding on the
first and last letter of the string because these are flanked by only
one other letter (Bouma, 1970, 1973; Estes, 1972; Estes, Allmeyer,
& Reder, 1976; Haber & Standing, 1969; Van der Heijden, 1992).
However, this explanation of the serial position function for letters
does not account for why the function changes for symbols and
non-letter shapes.
One possible account of the different serial position functions
found for letters and symbols is that reading-specific attentional
factors are at play with alphabetic stimuli. There is a long tradition
of research demonstrating attentional influences on the perception
of letters and words (e.g., Ducrot & Grainger, 2007; Heron, 1957).
On presentation of a centrally fixated string of letters attention
could be automatically drawn to the beginning of the stimulus,
hence facilitating processing of the first letter. Tydgat and
Grainger (2009) tested for the possible role of such attentional
biases in generating stimulus-specific serial position functions, by
testing letter targets embedded in strings of symbols and symbol
In languages that use alphabetical orthographies, the very first
stage of the reading process involves mapping visual features onto
representations of the component letters of the currently fixated
word (see Grainger, 2008, for a recent review). This rapid computation of a set of letter identities, likely performed in parallel, is
already a major challenge for the literate brain, given the limits in
visibility imposed by visual acuity and lateral interference (crowding) across neighboring letters. The challenge is all the greater for
the beginning reader, who must rapidly master the new art of
processing visual objects in close spatial proximity.
Strings of letters, even when they are meaningless and unpronounceable (e.g., HFKMT), are not processed in the same way as
strings of stimuli of similar visual complexity but that typically do
not occur in strings. This fact was first demonstrated using the
target search task, comparing serial position functions (search time
as a function of position in string) for letter stimuli versus symbol
stimuli or simple shapes. Search times for a target letter in a string
of letters show an approximate M-shaped serial position function
(i.e., shortest reaction times [RTs] for the first, third, and fifth
positions of a five-letter string). Symbol (e.g., ⌺, ␺) and shape
stimuli, on the other hand, show a U-shaped function with shortest
RTs for targets at the central position (on fixation) that increase as
Jonathan Grainger, Laboratoire de Psychologie Cognitive, Centre National de la Recherhce Scientifique and Aix-Marseille University; Ilse
Tydgat and Joanna Isselé, Department of Experimental Psychology, Ghent
University.
Ilse Tydgat is currently a research assistant for the National Fund for
Scientific Research, Flanders, Belgium.
Part of the research reported in this article was performed while Ilse
Tydgat and Joanna Isselé were on an Erasmus exchange between Ghent
University and the University of Provence. We thank Marine Salery for
running Experiment 4.
Correspondence concerning this article should be addressed to Jonathan
Grainger, Laboratoire de Psychologie Cognitive, Université de Provence, 3
pl. Victor Hugo, 13331, Marseille, France. E-mail: jonathan.grainger@
univ-provence.fr
673
674
GRAINGER, TYDGAT, AND ISSELÉ
targets embedded in strings of letters. They found that it was the
nature of the target, not the surrounding context, which determined
the form of the serial position function. This result also allowed
Tydgat and Grainger to reject a role for higher-order units (i.e.,
letter clusters or supra-letter features) as the basis of the different
serial position functions for letters and symbols. Finally, randomly
intermixing letter and symbol trials in an experiment had no effect
on the serial position function compared with blocked lists of
letters and symbols.
On the basis of their results, Tydgat and Grainger (2009) tentatively concluded that the different serial position functions found
for strings of letters and symbols might be related to differences in
the way crowding affects these two types of stimuli. More specifically, Tydgat and Grainger proposed that crowding effects might
be more limited in spatial extent for letter stimuli compared with
symbol stimuli, such that a single flanking stimulus would suffice
to generate close to maximum interference for symbols, but not for
letters. This would account for the superior performance at the first
and last positions for letter stimuli, but not for symbol stimuli.
According to Tydgat and Grainger, during reading acquisition, a
specialized system develops in order to optimize processing of
strings of letters. One key aspect of this optimization is a reduction
in the size of integration fields associated with location-specific
letter detectors that perform parallel letter identification (Grainger
& van Heuven, 2003).1 It is this reduction in the size of integration
fields, specific to stimuli that typically appear in strings (letters and
digits), that results in less crowding for such stimuli compared with
other types of visual stimuli such as symbols and geometric shapes
(see Figure 1).
According to Tydgat and Grainger’s (2009) account of the
different serial position functions for letters and symbols, one
should be able to observe differential crowding effects for letters
and symbols using a standard crowding manipulation. This prediction is illustrated in Figure 1. The hypothesized larger integration fields of symbol detectors (right panel of Figure 1) compared
with letter detectors (left panel of Figure 1) would result in greater
interference from flanking stimuli. Here we adopt an account of
crowding as integration of inappropriate feature information from
neighboring stimuli during target identification (Levi, 2008; Pelli
et al., 2007). The larger the integration field involved in identifying
a given target at a given location, the greater the number of
features from neighboring stimuli that can interfere in target identification.
Does crowding vary as a function of type of stimulus? Prior
research has demonstrated crowding effects for letter stimuli and
nonletter stimuli such as simple shapes using the standard crowding manipulation where accuracy to identify isolated stimuli presented left or right of fixation is compared with identification of
the same stimuli at the same location but flanked by two other
stimuli (see Levi, 2008, for review). However, to our knowledge,
no prior study has directly compared effects of crowding on letters
and symbols. Huckauf and Nazir (2007) examined the effects of
training on performance to letter and symbol stimuli flanked
by two characters, but did not test for differences across these two
types of stimuli, and did not include an isolated target condition.
Therefore, the first aim of the present study is simply to provide a
comparison of standard crowding effects in letters and symbols.
Experiment 1 tests letters and symbols with the classic dual flanker
paradigm and eccentric stimulus presentation. The exact same
stimuli as tested in our prior work on serial position functions were
used here for comparability, and the eccentricity of targets corresponds to the eccentricity of the first and last positions of the
centrally located five-character strings tested by Tydgat and
Grainger (2009). We also chose to use the two-alternative forcedchoice (2AFC) procedure used by Tydgat and Grainger, in an
attempt to equate overall performance level to letters and symbols
by removing possible influences of category size.
Summing up, the present experiments were designed as a test of the
crowding explanation of differences in serial position functions found
for letter and symbol stimuli put forward by Tydgat and Grainger
(2009). To do so, in Experiments 1 through 4 we tested exactly the
same letter and symbol targets at the same eccentricity as the first and
last positions of the five-letter strings used by Tydgat and Grainger,
using the same response procedure (2AFC in Experiments 1–3, barprobe partial report in Experiment 4), but with a standard crowding
manipulation (eccentric targets flanked by other characters). To respect the conditions typically used to investigate serial positions
functions, letter targets were always flanked by letters, and symbol
targets always flanked by symbols. Finally, Experiment 5 tested a new
set of letter and symbol stimuli, that were more closely matched in
terms of visual complexity, and applied a critical spacing manipulation that provides a more fine-grained measure of the spatial extent of
crowding (e.g., Chung, 2007; Pelli et al., 2007).
Experiment 1
Method
Participants. Twenty-four students of the University of Provence, Marseille, volunteered to participate in the experiment. Their
mean age was 20 years. Two were male, and all were native
speakers and readers of French and reported having normal or
corrected-to-normal vision.
Figure 1. Illustration of the different size of integration fields used to
identify letters and symbols at a given retinal location, hypothesized by
Tydgat and Grainger (2009). The larger size of integration fields for
symbol detectors predicts greater crowding effects for symbols than for
letters in a standard flanking experiment with eccentrically located targets.
This arises because more information from flanking stimuli falls in the
target’s integration field. Targets are the central character of three characters presented either left or right of fixation.
1
Tydgat and Grainger (2009) used the term receptive field to describe
the spatial extent of integration of visual information for letter and symbol
detectors. Here we adopt the more theoretically neutral terminology of Pelli
and colleagues (e.g., Pelli, Palomares, & Majaj, 2004), using the term
integration field to underline the scale-invariant property of these locationspecific receptive fields.
CROWDING AND STIMULUS TYPE
Stimuli and design. Stimuli were identical to the letter and
symbol stimuli tested by Tydgat and Grainger (2009). These were
nine uppercase consonants (B, D, F, G, K, N, L, S, or T) and nine
keyboard symbols (%, /, ?, @, }, ⬍, £, §, or ␮). The target was
presented either in isolation, or as the middle character in an array
of three characters. In the latter case, the target was flanked by two
different characters from the same category. Targets could appear
at two possible locations, left or right of fixation at an eccentricity
of 1.2° of visual angle. This gave a 2 (Target Type: letter or
symbol) ⫻ 2 (Crowding: isolated or flanked) ⫻ 2 (Visual Field:
left or right) factorial design. For the purposes of the forced-choice
task, each target character was pseudo-randomly paired with an
alternative character and this combination was maintained in the
different crowding and visual field conditions. The incorrect alternative presented for forced-choice was never a flanking character
of the target. Each letter or symbol stimulus was presented eight
times as a target (four times in isolation and four times between
two flankers, equally divided between the two visual fields),
giving a total of 144 trials. Each target stimulus also appeared eight
times as a flanker, and eight times as an incorrect alternative in the
forced-choice task. All factors were manipulated as withinparticipants factors, and all conditions were randomly mixed
within the experiment.
Procedure. The experiment was run inside a dimly lit room
and was controlled using E-Prime software (Psychology Software
Tools, Inc., www.pstnet.com/eprime). Participants were seated in
front of a computer screen at a viewing distance of approximately
60 cm. Stimuli were displayed in white on a black background, and
were presented in 18-point Courier New font. Each character
subtended 0.44° of visual angle, and the center-to-center spacing
was 0.6°. Stimuli were exposed at 92 cd/m2 luminance on a
background luminance of 4 cd/m2.
Participants first saw instructions and practiced the task during
20 trials, randomly chosen out of the stimulus list, to become
familiarized with the stimuli and the procedure. Each trial began
with two vertical fixation bars above and below the center of the
screen. After 1,012 ms, the fixation bars disappeared and the
stimulus (that consisted of either one or three characters) immediately appeared for a duration of 100 ms. This was followed by a
backward mask in which the stimulus characters were replaced by
hash marks (single hash mark for isolated targets, three hash marks
for flanked targets). The mask was accompanied by two characters, presented above and below the position of the target. Participants had to decide which of these two characters had been
present in the corresponding position of the preceding array (twoalternative forced-choice, 2AFC). They were asked to respond as
accurately as possible by pressing one of two keys on the keyboard. They had to choose either the upward arrow key (for the
alternative above), or the downward arrow key (for the alternative
below). After the response, the screen was cleared, and the two
fixation bars appeared again on the screen, introducing the next
trial. The participants could take a short break after half of the
trials. The experiment lasted approximately 15 min.
Results
Mean accuracies are presented in Table 1. A 2 (Target Type) ⫻
2 (Crowding) ⫻ 2 (Visual Field) analysis of variance (ANOVA)
was performed on the accuracy data. Performance to letter stimuli
675
Table 1
Percentage Correct in the Two-Alternative Forced-Choice
Procedure of Experiment 1 for Letters and Symbols in the Left
and Right Visual Field, and Either in Isolation or Flanked by
Two Other Characters
Accuracy (%)
Letters
Symbols
Visual field
Isolated
Two
flankers
Isolated
Two
flankers
Left
Right
91.4 (1.9)
92.1 (2.1)
74.1 (2.4)
78.7 (2.2)
89.4 (2.0)
90.5 (1.7)
67.8 (1.9)
66.7 (2.6)
Note.
SEs in parentheses.
was slightly higher than performance to symbol stimuli, and this
main effect of Target Type was significant in the participant
analysis, F1(1, 23) ⫽ 21.25, p ⬍ .001. There was a main effect of
Crowding, F1(1, 23) ⫽ 159.84, p ⬍ .001, F2(1, 16) ⫽ 81.60, p ⬍
.001, with mean performance for targets in isolation at 90.1%
compared with 71.8% for flanked targets. Most important, there
was an interaction between Target Type and Crowding which was
clearly significant in the analysis by participants, F1(1, 23) ⫽
11.83, p ⬍ .01, and approached significance in the analysis by
items F2(1, 16) ⫽ 2.99, p ⫽ .103. A significant crowding effect
was found separately for symbols, F1(1, 23) ⫽ 156.44, p ⬍ .0001,
F2(1, 16) ⫽ 57.92, p ⬍ .001 and for letters, F1(1, 23) ⫽ 67.82, p ⬍
.001, F2(1, 16) ⫽ 26.67, p ⬍ .001, but the size of this effect was
greater for symbols than for letters, with respectively 22.7% and
15.4% decrease in performance in the presence of flankers. There
was no effect of Visual Field, and no interactions with this factor.
Discussion
The results of Experiment 1 suggest that symbol targets are
more affected by the presence of flanking stimuli than are letter
targets. This result is all the more important given the fact that
performance to isolated symbols and letters was very similar.
Performance to both types of stimuli dropped in the presence of
two flankers, independently of visual field, but this drop in performance was greater for symbol stimuli. This result is in line with
Tydgat and Grainger’s (2009) prediction that crowding effects
should be greater for symbols than for letters.
However, Tydgat and Grainger’s account of the differences in
the serial position functions for letters and symbols, leads to one
more precise prediction concerning crowding effects in letters and
symbols. That is that letter stimuli should show greater release
from crowding in the presence of a single flanking stimulus
compared with the standard dual flanker situation. This was proposed in order to account for the greater outer position advantage
(better identification of targets in the first and last positions in
strings) found for letters than for symbols. Due to the relatively
small integration fields of letter detectors, the presence of a single
flanking character would generate little interference (see Figure 1).
On the other hand, given their relatively large integration fields,
symbol stimuli would suffer extensive interference from a single
flanker. Experiment 2 tests this prediction by examining effects of
GRAINGER, TYDGAT, AND ISSELÉ
676
a single flanker on performance to letters and symbols. We expect
to exaggerate the difference in crowding effects for letters and
symbols compared with Experiment 1.
Experiment 2
flanker that are closer to the effects of double flankers than to the
isolated flanker condition, whereas we expect letter targets to show
crowding effects from a single flanker that are closer to the
isolated target condition than the double flanker condition. A
combined analysis of these conditions in Experiments 1 and 2
provides an initial investigation of this prediction.
Method
Participants. Twenty-four students of the University of Provence, Marseille, volunteered to participate in the experiment. Their
mean age was 21 years. Six were male, and all were native
speakers and readers of French and reported having normal or
corrected-to-normal vision. None had participated in the previous
experiment.
Stimuli, design, and procedure. The stimuli and procedure
were the same as in Experiment 1. The design was also similar,
except that the isolated and double flanker conditions were replaced by two single flanker conditions. The target was either
accompanied by a single character on the left side, or by a single
character on the right side.
Results
A 2 (Target Type) ⫻ 2 (Flanker Position) ⫻ 2 (Visual Field)
ANOVA was performed on the accuracy data. Percentage correct
for each condition is shown in Table 2. The main effect of Target
Type was significant, F1(1, 23) ⫽ 66.62, p ⬍ .001, F2(1, 16) ⫽
14.90, p ⬍ .01. Performance was lower for symbol targets than for
letter targets, with respectively 76.0% and 90.2% accuracy. There
were no significant main effects of Flanker Position (left vs. right)
or Visual Field, and there were no significant interactions.
Discussion
The results of Experiment 2 show that symbol stimuli are more
subject to interference from a single flanking stimulus than letter
stimuli, thus providing a further confirmation of the predictions of
Tydgat and Grainger (2009). According to Tydgat and Grainger,
letter stimuli would be less prone to interference from a single
flanking stimulus than would be symbol stimuli, and this would be
the cause of the greater release from crowding at the outer positions in strings for letter stimuli. On the basis of this account, we
expect symbol stimuli to show crowding effects from a single
Table 2
Percentage Correct in the Two-Alternative Forced-Choice
Procedure of Experiment 2 for Letters and Symbols in the Left
and Right Visual Field, and Either With a Flanker on the Left
or on the Right of the Target Character
Combined Analysis of Experiments 1 and 2
Results
The results of Experiments 1 and 2 are plotted together in Figure 2.
Two analyses were performed using a 2 (Target Type) ⫻ 2 (Crowding) ⫻ 2 (Visual Field) design. In the first analysis, the Crowding
factor contrasted the isolated target condition of Experiment 1 and
single flanker condition of Experiment 2. There were significant
main effects of Target Type, F1(1, 46) ⫽ 52.76, p ⬍ .001, F2(1,
16) ⫽ 6.05, p ⬍ .05 and Crowding, F1(1, 46) ⫽ 14.40, p ⬍ .001,
F2(2, 32) ⫽ 42.26, p ⬍ .01, and a significant Target Type ⫻
Crowding interaction, F1(1, 46) ⫽ 31.26, p ⬍ .001, F2(2, 32) ⫽
4.47, p ⬍ .05. In the second analysis, the Crowding factor contrasted the single flanker condition of Experiment 2 and the double
flanker condition of Experiment 1. Again, this analysis showed
significant main effects of Target Type, F1(1, 46) ⫽ 87.17, p ⬍
.001, F2(1, 16) ⫽ 6.05, p ⬍ .05, and Crowding, F1(1, 46) ⫽ 31.09,
p ⬍ .001, F2(2, 32) ⫽ 42.26, p ⬍ .01, and a significant interaction,
F1(1, 46) ⫽ 4.14, p ⬍ .05, F2(2, 32) ⫽ 4.47, p ⬍ .05.
Discussion
The combined analysis of Experiments 1 and 2 suggests that
letter stimuli indeed show the greatest release from crowding in the
presence of a single flanking stimulus compared with the standard
dual flanker situation. The difference between letter and symbol
targets is exaggerated in the single flanker condition of Experiment
2 (14%) compared with the double flanker condition of Experiment 1 (9%). This result is in line with the predictions derived
from Tydgat and Grainger’s (2008) account of different serial
position functions for letters and symbols. Identification of symbol
targets is greatly affected by the presence of a single flanking
stimulus, whereas letter identification is hardly affected at all. This
therefore explains why outer letters (first and last position in
strings) are identified better than interior letters, whereas symbols
at the outer positions in strings are not identified better than at
interior positions. Experiment 3 provides a further test of the
predicted differential effects of single flankers on letter and symbol identification, while reducing stimulus exposure duration in
order to remove any possible ceiling effects that might have
occurred in Experiments 1 and 2.
Accuracy (%)
Letters
Experiment 3
Symbols
Visual field
Left
flanker
Right
flanker
Left
flanker
Right
flanker
Left
Right
89.6 (1.8)
89.6 (1.8)
91.2 (2.7)
90.5 (1.6)
78.0 (2.6)
75.0 (2.2)
78.0 (2.8)
72.9 (3.0)
Note.
SEs in parentheses.
Method
Participants. Twenty students of the University of Provence,
Marseille, volunteered to participate in the experiment. None had
participated in Experiments 1 or 2. Their mean age was 22 years.
Five were male, and all were native speakers and readers of French
and reported having normal or corrected-to-normal vision.
CROWDING AND STIMULUS TYPE
677
Figure 2. Percent correct in the two-alternative forced-choice procedure of Experiments 1 and 2 for letters and
symbols in isolation, flanked by one character (averaged across left and right flankers) or flanked by two
characters, in the left visual field (LVF) and right visual field (RVF). Stimulus duration was 100 ms, and vertical
bars represent SEs.
Stimuli, design, and procedure. The design was a combination of Experiments 1 and 2, such that the Crowding manipulation
now involved four levels. Targets were presented in isolation, with
a single flanker to the left, a single flanker to the right, or with two
flankers (one flanker on each side of the target). Stimuli were
presented for 66 ms. Participants received 288 experimental trials.
Otherwise the procedure was identical to the previous experiments.
Results
The results of Experiment 3 are shown in Figure 3. A 2 (Target
Type) ⫻ 4 (Crowding) ⫻ 2 (Visual Field) ANOVA was performed
on 2AFC accuracy in each condition. There was a main effect of
Target Type in the participant analysis, F1(1, 19) ⫽ 51.47, p ⬍ .001,
but not in the item analysis. There was also a main effect of Crowding, F1(3, 57) ⫽ 29.84, p ⬍ .001, F2(3, 48) ⫽ 24.88, p ⬍ .001, and
of Visual Field, F1(1, 19) ⫽ 6.11, p ⬍ .05, F2(1, 16) ⫽ 9.22, p ⬍ .01.
The interaction between Target Type and Crowding was significant,
F1(3, 57) ⫽ 6.31, p ⬍ .01, F2(3, 48) ⫽ 3.60, p ⬍ .05.
Follow-up analyses were performed separately for letters and
symbols. A 4 (Crowding) ⫻ 2 (Visual Field) ANOVA for letters
revealed a significant main effect of Crowding, F1(3, 57) ⫽ 20.08,
p ⬍ .001, F2(3, 24) ⫽ 19.09, p ⬍ .001. Contrast analyses (col-
lapsing data across the two single flanker conditions) further
showed that performance to isolated letters was significantly
higher than performance to single flanked letters, F1(1, 19) ⫽ 9.94,
p ⬍ .01, F2(1, 16) ⫽ 5.97, p ⬍ .05, and performance to single
flanked letters was significantly higher than to double flanked
letters, F1(1, 19) ⫽ 25.03, p ⬍ .001, F2(1, 16) ⫽ 13.71, p ⬍ .01.
There was also a main effect of Visual Field, F1(1, 19) ⫽ 5.26, p ⬍
.05, F2(1, 8) ⫽ 5.29, p ⬍ .05, and a significant Crowding x Visual
Field interaction in the participant analysis, F1(3, 57) ⫽ 4.28, p ⬍
.01. As can be seen in Figure 3, this interaction is driven by the
different effects of the single flanker condition as a function of
visual field. Indeed, when the analysis was restricted to these two
crowding conditions (a 2 ⫻ 2 ANOVA), the Crowding ⫻ Visual
Field interaction was still significant in the participant analysis,
F1(1, 19) ⫽ 12.35, p ⬍ .01, and now approached significance in
the item analysis, F2(1, 8) ⫽ 4.23, p ⫽ .07. Identification of target
letters in the left visual field (LVF) was affected by flanker
position (left vs. right), but not in the right visual field (RVF; see
Figure 4).
The separate analysis (4 ⫻ 2 ANOVA) for symbols only revealed a significant main effect of Crowding, F1(3, 27) ⫽ 10.78,
p ⬍ .001, F2(3, 24) ⫽ 11.95, p ⬍ .001. Performance to single
Figure 3. Percent correct in the two-alternative forced-choice procedure of Experiment 3 for letters and
symbols in isolation, flanked by one character (averaged across left and right flankers), or two characters with
one character at each side, in the left visual field (LVF) and right visual field (RVF). Stimulus duration was 66
ms, and vertical bars represent SEs.
678
GRAINGER, TYDGAT, AND ISSELÉ
Figure 4. Percent correct in the two-alternative forced-choice procedure of Experiment 3 for letters and
symbols flanked by one character on the left side, or flanked by one character on the right side, in the left visual
field (LVF) and right visual field (RVF). Stimulus duration was 66 ms, and vertical bars represent SEs.
flanked symbols (collapsing across flanker position) did not differ
significantly from performance to double flanked symbols, but was
significantly lower than performance to isolated symbols, F1(1,
19) ⫽ 24.58, p ⬍ .001, F2(1, 16) ⫽ 20.75, p ⬍ .001.
Discussion
The results of Experiment 3 provide a direct replication of the
combined results of Experiments 1 and 2. Symbol targets were
more affected by crowding than were letter stimuli, and this was
particularly evident in the single flanker condition. For symbol
targets, performance with a single flanker did not differ significantly from the double flanker condition, while for letter targets
performance in the single flanker condition was closer to the
isolated target condition. This result is in line with Tydgat and
Grainger’s (2009) prediction that symbol targets presented in
strings suffer practically maximum interference from a single
flanking stimulus. Letters, on the other hand, show about half as
much interference from a single flanker compared with double
flankers.
One particularity of the results with letter targets is that there is
a distinct asymmetry in the effects of flanker position (left vs.
right) in the single flanker condition, and this asymmetry only
appeared for letter targets presented in the LVF. No such asymmetry was apparent in Experiment 2, but this might be due to the
overall higher level of performance to letter stimuli in that experiment (a ceiling effect). Indeed, prior research examining effects of
single flanker stimuli with letter targets has shown such an asymmetrical effect, with stronger interference arising when the flanker
appears in the peripheral position (to the left for stimuli presented
to the left, and to the right for stimuli presented to the right),
compared to flankers positioned on the foveal side of targets
(Banks, Bachrach, & Larson, 1977; Banks, Larson, & Prinzmetal,
1979; Bouma, 1970; Chastain, 1982; Krumhansl & Thomas, 1976,
1977; White, 1981). There is also evidence from prior research that
such asymmetrical crowding effects are stronger in the LVF than
the RVF, at least in studies using letter stimuli (Krumhansl &
Thomas, 1977; White, 1976). This was clearly the case in Experiment 3, where flanker position only affected performance to letter
targets in the LVF. We return to discuss the relevance of this
particular finding with respect to serial position functions for
letters and symbols in the General Discussion.
Experiment 4
One possible limitation of the 2AFC task used in Experiments
1–3 is that it does not allow for location errors when a participant
has identified the target but incorrectly reports the identity of a
flanking stimulus (the flanking stimuli were never presented as a
possible response). It is therefore possible that a partial report
procedure, including such location errors, might produce a different pattern of results. Experiment 4 therefore tests the generality of
the findings of Experiment 3 using a partial report procedure. The
same stimulus exposure as Experiment 3 is used, plus a shorter
duration, given that Tydgat and Grainger (2009) found that the
partial report procedure generated a higher level of performance
than 2AFC.
Method
Participants. Forty-one students of the University of Provence, Marseille, volunteered to participate in the experiment. None
had participated in the previous experiments. Their mean age was
23 years. Seven were male, and all were native speakers and
readers of French and reported having normal or corrected-tonormal vision.
Stimuli, design, and procedure. The procedure was identical
to Experiment 3 except that the 2AFC task was replaced by a
partial report bar-probe task, and presentation duration of the
stimulus was either 50 ms or 66 ms (manipulated betweenparticipants). The backward mask was accompanied by two hyphen marks positioned above and below one of the positions in the
string (hence mimicking the position of the two alternative letters
in the 2AFC procedure). Participants were asked to identify the
character they had seen at this position by pressing one of 18
marked response keys (9 letters and 9 symbols) on the keyboard
placed in front of them.
Results
The results of Experiment 4 are shown in Figures 5 and 6. A 2
(Stimulus Duration) ⫻ 2 (Target Type) ⫻ 2 (Visual Field) ⫻ 4
CROWDING AND STIMULUS TYPE
679
Figure 5. Percent correct in Experiment 4 (partial report bar-probe task) for letters and symbols in isolation,
flanked by one character (averaged across left and right flankers), or double flanked with one character at each
side, in either the left visual field (LVF) or right visual field (RVF). Vertical bars represent SEs.
(Crowding) ANOVA was performed on identification accuracy in
each condition. There were main effects of Stimulus Duration,
F1(1, 39) ⫽ 5.23, p ⬍ .05, F2(1, 16) ⫽ 97.92, p ⬍ .001, Target
Type, F1(1, 39) ⫽ 203.72, p ⬍ .001, F2(1, 16) ⫽ 22.71, p ⬍ .001,
Visual Field, F1(1, 39) ⫽ 8.67, p ⬍ .01, F2(1, 16) ⫽ 2.04, p ⬍
.1, and Crowding, F1(3, 117) ⫽ 195.11, p ⬍ .001, F2(3, 48) ⫽
98.80, p ⬍ .001. Except for Stimulus Duration ⫻ Target Type,
F1(1, 39) ⫽ 1.66, p ⫽ .21, F2(1, 16) ⫽ 8.24, p ⬍ .05, none of the
interactions with Stimulus Duration approached significance.2 The
critical interaction between Target Type and Crowding was significant in the participants analysis, F1(3, 117) ⫽ 6.20, p ⬍ .001,
and approached significance in the item analysis, F2(3, 48) ⫽ 2.30,
p ⬍ .1, and Crowding significantly interacted with Visual Field,
F1(3, 117) ⫽ 43.86, p ⬍ .001, F2(3, 48) ⫽ 12.58, p ⬍ .001. As can
be seen in Figures 5 and 6, the pattern of crowding effects was
different for letters and symbols, and effects of single flankers
were different in the LVF and RVF. As in the previous experiments, separate analyses were performed for letter and symbol
targets to investigate these interactions.
A 4 (Crowding) ⫻ 2 (Visual Field) ANOVA for letters revealed
a significant main effect of Crowding, F1(3, 120) ⫽ 91.99, p ⬍
.001, F2(3, 24) ⫽ 45.13, p ⬍ .001. Contrast analyses (collapsing
data across the two single flanker conditions) further showed that
performance for isolated letters was significantly higher than for
single flanked letters, F1(1, 40) ⫽ 20.61, p ⬍ .001, F2(1, 8) ⫽
8.03, p ⬍ .05, and performance for single flanked letters was
significantly higher than for double flanked letters, F1(1, 40) ⫽
121.62, p ⬍ .001, F2(1, 8) ⫽ 51.05, p ⬍ .001. There was a
significant interaction between Crowding and Visual Field, F1(3,
120) ⫽ 17.72, p ⬍ .001, F2(3, 24) ⫽ 6.39, p ⬍ .01. Figure 6
reveals that this interaction was driven by differences in the single
flanker condition as a function of visual field. Contrast analyses
revealed an asymmetrical effect of flanker position in the LVF,
with greater interference from a left flanker (peripheral) compared
with a right flanker (foveal), F1(1, 40) ⫽ 32.15, p ⬍ .001, F2(1,
8) ⫽ 58.88, p ⬍ .001. In the RVF, there was a trend toward greater
interference from a right flanker (peripheral) compared with a left
flanker (foveal), F1(1, 40) ⫽ 3.10, p ⬍ .1, F2(1, 8) ⫽ 7.17,
p ⬍ .05.
The separate analysis (4 ⫻ 2 ANOVA) for symbols also revealed a significant main effect of Crowding, F1(3, 120) ⫽ 157.32,
p ⬍ .001, F2(3, 24) ⫽ 57.25, p ⬍ .001. Performance for isolated
symbols was significantly higher than for single flanked symbols,
F1(1, 40) ⫽ 86.33, p ⬍ .001, F2(1, 8) ⫽ 27.59, p ⬍ .001, and
performance for single flanked symbols was significantly higher
than for double flanked symbols, F1(1, 40) ⫽ 203.75, p ⬍ .001,
F2(1, 8) ⫽ 116.58, p ⬍ .001. There was a significant interaction
between Crowding ⫻ Visual Field, F1(3, 120) ⫽ 27.19, p ⬍ .001,
F2(3, 24) ⫽ 6.77, p ⬍ .01. Figure 6 reveals that this interaction
reflects an asymmetrical effect of flanker position in both the left
and right visual fields. Flankers in the outer (peripheral) position
generated more interference than flankers in the inner (foveal)
position for symbol targets in the left and right visual field, F1(1,
40) ⫽ 45.56, p ⬍ .001, F2(1, 8) ⫽ 34.20, p ⬍ .001, and F1(1,
40) ⫽ 13.29, p ⬍ .001, F2(1, 8) ⫽ 4.99, p ⬍ .1, respectively.
Error analyses. In the partial report task used in Experiment
4, errors can be classified as item errors—when the erroneous
response is not one of the flanking characters, or location errors—
when one of the flanking characters is reported instead of the
target. A 2 (Error Type) ⫻ 2 (Target Type) ⫻ 2 (Visual Field) ⫻
3 (Crowding) was performed. The isolated crowding condition was
excluded from this analysis since errors in this condition were
always item errors. The results broken down by Error Type are
reported in Table 3. Of most relevance here is the significant
2
To verify that the different pattern of crowding effects for letters and
symbols was not being driven by a ceiling effect for letter targets, we
performed an ANOVA introducing the performance level of our participants as a between participants variable. Participants were divided into two
equal groups according to their average performance in the experiment (20
participants with the lowest performance in one group, and 20 participants
with the highest performance in the other group, with 1 participant removed). Performance Level interacted with Crowding, F1(3, 114) ⫽ 5.06,
p ⬍ .01, F2(3, 48) ⫽ 6.91, p ⬍ .001, and there was a significant Performance Level ⫻ Target Type ⫻ Crowding interaction, F1(3, 114) ⫽ 16.56,
p ⬍ .001, F2(3, 48) ⫽ 12.92, p ⬍ .001. A separate ANOVA for the
low-performance group was performed in order to remove the ceiling
effect for letter targets that was driving these interactions. The results of the
low-performance group largely mimicked the pattern of the whole group,
with main effects of Target Type, F1(1, 19) ⫽ 107.42, p ⬍ .001, F2(1,
16) ⫽ 25.86, p ⬍ .001, and Crowding, F1(3, 57) ⫽ 101.22, p ⬍ .001, F2(3,
48) ⫽ 71.44, p ⬍ .001, and the critical Target Type ⫻ Crowding interaction was significant, F1(3, 57) ⫽ 10.62, p ⬍ .001, F2(3, 48) ⫽ 4.38, p ⬍
.001.
GRAINGER, TYDGAT, AND ISSELÉ
680
Figure 6. Percent correct in Experiment 4 (partial report bar-probe task), for letters and symbols flanked by one
character on the left side, or flanked by one character on the right side, in either the left visual field (LVF) or
right visual field (RVF). Vertical bars represent SEs.
interaction in the by participants analysis between Error Type,
Target Type, Visual Field, and Crowding, F1(2, 76) ⫽ 3.80, p ⬍
.05, F2(2, 32) ⬍ 1. Follow-up analyses examined the results
separately for each type of error.
A 2 (Target Type) ⫻ 2 (Visual Field) ⫻ 3 (Crowding) ANOVA
on location errors revealed significant main effects of Target
Type, F1(1, 40) ⫽ 53.16, p ⬍ .001, F2(1, 16) ⫽ 10.42, p ⬍ .01,
and Crowding, F1(2, 80) ⫽ 153.72, p ⬍ .001, F2(2, 32) ⫽ 124.49,
p ⬍ .001. The pattern of Crowding effects was quite similar for
letter and symbol targets, with significant main effects of Crowding for both letters, F1(2, 80) ⫽ 78.95, p ⬍ .001, F2(2, 16) ⫽
58.78, p ⬍ .001, and symbols, F1(2, 80) ⫽ 99.32, p ⬍ .001, F2(2,
16) ⫽ 66.74, p ⬍ .001, and a significant Visual Field ⫻ Crowding
interaction for both types of target, F1(2, 80) ⫽ 10.37, p ⬍ .001,
F2(2, 16) ⫽ 1.15, p ⫽ .34, and F1(2, 80) ⫽ 11.23, p ⬍ .001, F2(2,
16) ⫽ 4.27, p ⬍ .05, respectively. For letters, this interaction
reflected an asymmetrical effect of flanker position in the LVF—
with greater interference from a left flanker (peripheral) compared
with a right (foveal) flanker, F1(1, 40) ⫽ 15.80, p ⬍ .001, F2(1,
8) ⫽ 39.76, p ⬍ .001—and no asymmetrical effect in the RVF. For
symbol targets, flankers in the outer (peripheral) position generated more interference than flankers in the inner (foveal) position
in both the left and right visual field, F1(1,40)⫽5.01, p ⬍ .05,
F2(1, 8) ⫽ 16.29, p ⬍ .01, and F1(1, 40) ⫽ 17.04, p ⬍ .001, F2(1,
8) ⫽ 9.78, p ⬍ .05, respectively.
A 2 (Target Type) ⫻ 2 (Visual Field) ⫻ 3 (Crowding) ANOVA on
item errors revealed significant main effects of Target Type, F1(1,
40) ⫽ 120.26, p ⬍ .001, F2(1, 16) ⫽ 25.83, p ⬍ .001, Visual Field,
F1(1, 40) ⫽ 14.91, p ⬍ .001, F2(1, 16) ⫽ 4.98, p ⬍ .05, and
Crowding, F1(2, 80) ⫽ 31.13, p ⬍ .001, F2(2, 32) ⫽ 9.37, p ⬍ .001.
The significant two-way interaction between Target Type and Crowding, F1(2, 80) ⫽ 8.70, p ⬍ .001, F2(2, 32) ⫽ 3.01, p ⬍ .1, reflects the
fact that letter targets showed a large difference between the single
and double flanker conditions, F1(1, 40) ⫽ 47.62, p ⬍ .001, F2(1,
8) ⫽ 15.69, p ⬍ .01, whereas this difference was not significant for
symbols. There was a significant interaction between Visual Field and
Crowding, F1(2, 80) ⫽ 23.02, p ⬍ .001, F2(2, 32) ⫽ 8.65, p ⬍ .001,
which did not interact with Target Type. For both letters and symbols
this interaction reflects an asymmetrical effect of flanker position in
the LVF—with greater interference from a left (peripheral) flanker
compared with a right (foveal) flanker, F1(1, 40) ⫽ 19.82, p ⬍ .001,
F2(1, 8) ⫽ 22.15, p ⬍ .01 for letters, and F1(1, 40) ⫽ 29.66, p ⬍ .001,
F2(1, 8) ⫽ 16.61, p ⬍ .01 for symbols, while there was no effect of
flanker position in the RVF for either letters or symbols.
Discussion
The results of Experiment 4 provide a straightforward replication of the results of Experiment 3 with a partial report bar-probe
procedure replacing 2AFC. Crowding was once again found to
Table 3
Results Broken Down by Type of Error in Experiment 4
Letters
Visual field
Left flanker
Right flanker
Left
Right
93.5 (1.6)
97.8 (0.6)
98.8 (0.5)
96.9 (0.9)
Left
Right
88.5 (2.2)
95.3 (1.2)
97.3 (0.8)
92.3 (2.1)
Symbols
Two flankers
Left flanker
Location errors
79.9 (2.5)
89.7 (1.4)
76.6 (2.7)
93.0 (1.6)
Item errors
78.6 (2.4)
58.1 (3.2)
84.3 (1.8)
69.2 (3.1)
Right flanker
Two flankers
93.2 (1.2)
85.5 (1.3)
67.2 (2.1)
70.1 (2.1)
71.3 (3.2)
68.3 (2.9)
61.0 (2.7)
66.8 (2.8)
Note. Data are percentage correct target identification when counting only location errors (upper panel) or only item errors (lower panel), with SEs in
parentheses.
CROWDING AND STIMULUS TYPE
have a different effect on the identification of letters compared
with symbol stimuli. However, in Experiment 4 the critical Crowding ⫻ Target Type interaction was mostly driven by the differential effects of single flankers compared with the isolated target
condition. In line with the results obtained with 2AFC, symbol
targets were more affected by the presence of a single flanker than
were letter targets. This result was particularly marked in the item
errors, when a character other than the target or the flankers was
reported. When counting only such item errors, there was no
difference in performance to symbol targets with single or double
flankers, while letter targets showed significantly more item errors
with double flankers than single flankers. These results provide
further support for Tydgat and Grainger’s (2009) hypothesis that
the outer position advantage found for letters compared with
symbols in target-in-string identification is because of differential
crowding effects from single flankers.
There were some other notable differences in the pattern of
results obtained in the bar-probe partial report procedure of Experiment 4 compared with 2AFC in Experiment 3 (remember that
the experiments were otherwise identical except for the use of an
additional shorter stimulus duration). First, performance to isolated
symbol stimuli was significantly lower than to isolated letters in
Experiment 4. As noted in our presentation of the 2AFC procedure, this might be because the 2AFC procedure reduces possible
differences due to category set size, which would endow an advantage for the closed-set category of letter stimuli relative to the
open-set category of symbols. Second, Experiment 4 showed a
significant decrement in performance in the double flanker condition compared with the single flanker condition for both letters and
symbols, whereas in Experiment 3 this was not the case for symbol
stimuli. This pattern actually fits very well with the variation in
serial position functions found by Tydgat and Grainger (2009) as
a function of type of report (2AFC vs. partial report). The serial
position function for symbol stimuli became slightly W-shaped
(with a significant quadratic component) in the partial report
procedure, whereas a systematic inverted U-shaped function (with
no quadratic component) was always found with the 2AFC procedure. The results of Experiment 4 of the present study suggest
that this could be because of the increased effect of number of
flankers (single vs. double) for symbol stimuli in the partial report
procedure compared with 2AFC. The 2AFC task removes location
errors (since the two alternatives were never a flanking character),
and it was such location errors that were found to be driving the
difference between the single and double flanker conditions for
symbol targets. Finally, the effects of flanker position in the single
flanker condition with symbol targets showed a different pattern in
Experiment 4. While letter targets continued to show stronger
interference from peripheral flankers in the LVF, symbol targets
now showed this asymmetric effect in both visual fields (where
there was no effect of flanker position in Experiment 3).
Experiment 5
The differential effects of crowding for letters and symbols
found in Experiments 1 through 4 could be driven by differences
in visual complexity across these two types of stimuli. Although
Pelli et al. (2007) have argued that crowding should not depend on
anything else but stimulus eccentricity, there is some evidence for
an influence of stimulus complexity on crowding (e.g., Põder,
681
2008). Therefore, to test for a possible role of stimulus complexity
in Experiments 1 through 4, we first calculated the complexity of
all individual letters and symbols used in these experiments. To
this end, we computed the perimetric complexity and the stroke
frequency of each of our letter and symbol stimuli (Pelli, Burns,
Farell, & Moore-Page, 2006). Because we are comparing small
subsets of stimuli in our experiments, instead of whole alphabets as
in Pelli et al. (2006), we adopted the stroke frequency calculation
proposed by Zhang, Zhang, Xue, Liu, and Yu (2007) for Chinese
characters. This involves drawing multiple horizontal, vertical and
oblique slices through each stimulus and counting the number of
the times these strokes cross a line in the stimulus, giving a mean
stroke frequency per stimulus. We adapted their method for our
letters and symbols using 3 horizontal and 3 vertical slices. These
three slices were drawn at respectively 25%, 50%, and 75% of the
total width and height. T-tests revealed that the letters and symbols
used in Experiments 1 through 4 did not differ significantly with
respect to either stroke frequency ( p ⫽ .48) or perimetric complexity ( p ⫽ .28).3
As a further test of any possible role for stimulus complexity,
Experiment 5 uses a new set of letter and symbol stimuli selected
to be more closely matched on these measures of visual complexity. Furthermore, Experiment 5 applies a target-flanker spacing
manipulation in order to measure critical spacing values for letter
and symbol stimuli.
Method
Participants. Twenty-nine students of the University of Provence, Marseille, volunteered to participate in the experiment. None
participated in any of the prior experiments. Their mean age was
22 years. Fourteen were female, and all were native speakers and
readers of French and reported having normal or corrected-tonormal vision.
Stimuli and design. The experiment consisted of 2 (Target
Type) ⫻ 2 (Visual Field) ⫻ 5 (Spacing) conditions. All variables
were manipulated within-subjects. Ten letter and ten symbol targets were selected on the basis of their scores on two measures of
visual complexity. The letter stimuli were lowercase consonants
(c, f, g, h, m, s, p, v, w, z) and the symbol set contained both
keyboard symbols and non-keyboard symbols (兺, ⬍, ⌿, &, @, £,
3
To provide a further test of the role of visual similarity in driving the
different effects of crowding for letters and symbols we selected a subset
of 7 letters (D, F, G, L, N, S, T) and 7 symbols (%, ?, @, ⬍, £, §, ␮) that
were very closely matched on our measures of visual complexity. We
chose to re-analyze the results of Experiment 3 with this subset of stimuli,
since this Experiment was not subject to possible ceiling effects, and the
subset of stimuli for re-analysis produced higher levels of performance
than the complete set. A 2 (Target Type) ⫻ 4 (Crowding) ANOVA was
conducted by participants. Overall accuracy was the same for symbols and
letters (74%). There was a main effect of Crowding, F(3, 57) ⫽ 33.49, p ⬍
.001, and a significant Target Type ⫻ Crowding interaction, F(3, 57) ⫽
5.50, p ⬍ .01. Follow-up analyses revealed a significant difference between the isolated and the single flanker condition only for symbols, F(1,
19) ⫽ 15.66, p ⬍ .001, and a significant difference between the single
flanker and two flankers conditions only for letters, F(1, 19) ⫽ 7.84, p ⬍
.05. This analysis establishes different effects of crowding for letters and
symbols in conditions where overall performance is matched, and both
ceiling and floor effects are avoided.
682
GRAINGER, TYDGAT, AND ISSELÉ
⬁, ␭, ⌸, 么). We computed perimetric complexity and mean stroke
frequency for both sets of stimuli (see description of these measures in the preceding section). The two types of stimuli did not
differ significantly on either of these measures ( p ⬎ .75). Targetto-flanker separation (Spacing) was measured as the center-tocenter distance in visual angle. Each target appeared in the five
spacing conditions and in both visual fields. Target stimuli were
always the central character of the trigram. Two characters from
the same stimulus category as the target were randomly assigned to
each target and functioned as flankers. Trials were divided into two
experimental blocks such that each target appeared at all five
spacing conditions in each block and either in the right or left
visual field (half of the targets in the RVF and half in the LVF).
Visual field location of each target was therefore counterbalanced
across blocks, and presentation of the trials was randomized within
each block, with a total of 100 trials per block.
Procedure. Viewing distance, font, luminance, stimulus size
and fixation duration were identical to the previous experiments.
After receiving the instructions, participants completed a practice
block of 20 trials. Subsequently, the experimental phase started.
Stimuli were presented in the right or left visual field at an
eccentricity of 3°. On the basis of a pilot study conducted with 20
participants, stimulus duration was set at 80 ms (well below
durations that would allow eye movements to the target), and the
target-flanker separations were chosen to be 0.5°, 0.9°, 1.3°, 1.7°
and 2.1°. Stimuli were presented in black on a white background
at the same luminance values as the previous experiments, and
were followed immediately by a blank screen, which lasted until a
response was triggered. Participants responded to the target by
choosing the corresponding key on the keyboard among the 20
alternatives (10 letters and 10 symbols). They were instructed to
respond on every trial, and the next trial began after participants’
response. The participants could take a short break at the end of the
first block. The experiment lasted approximately 15 min.
Results
Two participants were excluded from the analysis because of
failure to follow the instructions. A 2 (Target Type) ⫻ 2 (Visual
Field) ⫻ 5 (Spacing) ANOVA was performed on the data for
identification accuracy in each condition. The main effect of
Target Type was significant in the participant analysis, F1(1, 26) ⫽
17.32, p ⬍ .001, F2(1, 18) ⫽ 0.73, p ⫽ .40, reflecting a higher
overall accuracy for letters (82.6%) than for symbols (76.7%).
There was a main effect of Spacing, F1(4, 104) ⫽ 172.33, p ⬍
.001, F2(4, 74) ⫽ 34.59, p ⬍ .001. Mean accuracies ranged from
58.9% in the smallest spacing condition to 91.3% in the largest
spacing condition. There was a main effect of Visual Field, F1(1,
26) ⫽ 30.10, p ⬍ .001, F2(1, 18) ⫽ 7.36, p ⬍ .05, with higher
levels of performance for targets presented in the RVF. Most
important is that the Target Type ⫻ Spacing interaction was
clearly significant in the participant analysis, F1(4, 104) ⫽ 10.70,
p ⬍ .001 and marginally significant in the item analysis, F2(4,
72) ⫽ 2.06, p ⬍ .10. As can be seen in Figure 7, the interaction
reflects the decreasing difference in identification scores for letters
and symbols as spacing increases. None of the other interactions
were significant (Table 4).
Figure 7. Proportion correct target identification plotted as a function of target-to-flanker spacing (visual angle
in degrees), in both the left visual field (LVF, left panel) and the right visual field (RVF, right panel) in
Experiment 5. Fitted curves are cumulative-Gaussian functions. Critical spacing is defined as the target-toflanker spacing at which the proportion correct is equal to 0.75 (indicated with dotted lines and values shown
above the x-axis). Vertical bars represent SEs.
CROWDING AND STIMULUS TYPE
Table 4
Percentage Correct Identification of Letter and Symbol Targets
in Experiment 5 for the Different Target-to-Flanker Spacing
Conditions and for Both Visual Fields
Accuracy (%)
LVF
Letters
Symbols
Letters
Symbols
0.5
0.9
1.3
1.7
2.1
66.8 (2.5)
74.7 (2.5)
78.9 (2.6)
84.5 (2.5)
90.1 (2.2)
47.8 (3.1)
73.8 (2.9)
79.4 (2.7)
84.9 (2.4)
85.2 (2.6)
68.3 (3.0)
83.3 (2.9)
90.5 (1.9)
92.5 (1.7)
96.6 (1.0)
52.6 (2.7)
74.7 (2.7)
85.2 (2.7)
89.9 (1.7)
93.2 (1.4)
Note. LVF ⫽ left visual field; RVF ⫽ right visual field. SEs in parentheses.
Critical spacing values for letters and symbols were calculated
by fitting a cumulative Gaussian function to the proportion correct
scores as a function of target-flanker distance for each participant
and each condition of the Target Type and Visual Field factors. We
then calculated the value for target-to-flanker spacing that corresponded to a pre-defined criterion level of 75% accuracy. These
scores per Target Type and Visual Field were entered in a byparticipant ANOVA. The critical spacing was significantly smaller
for letters (0.82°) than for symbols (1.12°), F(1, 26) ⫽ 5.93, p ⬍
.05. Furthermore, there was a significant main effect of Visual
Field, F(1, 26) ⫽ 6.65, p⬍ .05, reflecting a larger critical spacing
in the LVF (1.12°), compared with the RVF (0.83°). Although
Figure 7 reveals that the difference between letters and symbols is
numerically greater in the RVF, the interaction between Target
Type and Visual Field was not significant.
Discussion
Experiment 5 applied a critical spacing manipulation
whereby character triples were presented with varying levels of
intercharacter spacing in order to establish the spacing required
to attain a criterion level of correct identification of the central
target. Stimuli were presented at a fixed eccentricity (3°) along
the horizontal meridian in either the LVF or RVF. The results
are clear-cut. Critical spacing values were found to be significantly greater for symbols than for letters. Criterion levels of
letter identification were reached with a smaller inter-character
spacing compared with the spacing required to attain the same
level of performance with symbol stimuli. These results suggest
once again that crowding affects letters and symbols differently,
at least for the eccentricities and locations tested in the present
study.
General Discussion
The results of the present study can be summarized as follows.
(1)
(2)
This differential crowding effect was found with a
2AFC procedure in Experiments 1 through 3, a partialreport bar-probe procedure in Experiment 4, and using a
critical spacing manipulation in Experiment 5.
(3)
In Experiments 1 through 4, letter stimuli showed very
little interference from single flankers compared with
the isolated condition, while symbol targets always
showed sizeable interference from single flankers.
(4)
Visual field influences on crowding effects were seen
with letter targets and single flankers in Experiments 2
through 4, where identification of targets in the left
visual field was hurt more by a leftward flanker than a
rightward flanker.
(5)
Post-hoc analyses allowed us to exclude a possible role
for stimulus complexity in driving the observed differences in crowding effects for letters and symbols, and
this factor was explicitly controlled for in Experiment 5.
RVF
Spacing (°)
Flanking stimuli interfere more strongly in the identification of symbol targets than letter targets in a standard
crowding manipulation with eccentrically located targets (left and right of a central fixation point).
683
The experiments of the present study were designed to test the
predictions derived from a recent account of serial position functions found for letter and symbol stimuli (Tydgat & Grainger,
2009). In certain conditions, letter stimuli showed a clear advantage for the first and last positions of the string that was not found
with symbol stimuli. Because one classic interpretation of the
shape of the serial position function found for letters is in terms of
reduced crowding at these external positions (e.g., Bouma, 1970;
Estes, 1972), Tydgat and Grainger suggested that symbol stimuli
might be affected differently by crowding. More precisely, a single
flanking stimulus might be sufficient to generate almost maximal
crowding effects with such stimuli, hence the lack of reduction in
crowding effects at the outer positions. In other words, letter
stimuli benefit from a greater release from crowding at the outer
positions than do symbol stimuli.
The results of the present study confirm the predictions derived
from this “crowding” interpretation of the different serial position
functions found for letter and symbol stimuli. In all experiments,
target letters were found be more prone to crowding than were
symbol targets. This was found to be the case in Experiments 1
through 4 using the classic flanker paradigm and the same letter
and symbol stimuli as Tydgat and Grainger placed at the same
location as the first and last positions of the five-character strings
tested in their experiments. Post-hoc analyses strongly suggested
that the different effects of crowding for letters and symbols could
not be attributed to differences in stimulus complexity across these
two types of stimuli. Experiment 5 provided a further confirmation
of greater crowding with symbols than letters using a new set of
stimuli matched for visual complexity, and applying a critical
spacing manipulation with targets located at greater eccentricities
than in Experiments 1 through 4. Critical spacing was found to be
smaller for letters than for symbols, with letter targets being
identified more accurately than symbol targets at the lowest levels
of intercharacter spacing.
Relative to the predictions of Tydgat and Grainger (2009), it is
important to note that in Experiments 1 through 4, target letters
were found to be practically immune to the present of a single
flanker compared with the isolated target condition. Symbol
684
GRAINGER, TYDGAT, AND ISSELÉ
targets, on the other hand, were systematically affected by the
presence of a single flanker, with significantly lower levels of
identification compared with the isolated target condition. The
outer-position advantage found for letter stimuli but not for symbol
stimuli in the target-in-string identification paradigm would therefore be due to the reduction in crowding from single flankers at the
outer positions in strings that occurs more for letters than for
symbols.
Mechanisms of Crowding
The fact that crowding was found to be affected by stimulus
type in the present study would appear to contradict purely
bottom-up accounts of this phenomenon (e.g., Pelli & Tillman,
2008; Pelli et al., 2004, 2007), and be more in line with accounts
that appeal to top-down mechanisms such as spatial attention (e.g.,
He, Cavanagh, & Intriligator, 1996; Intriligator & Cavanagh,
2001). However, our preferred interpretation of the present results
appeals uniquely to bottom-up mechanisms, whose operation can
vary as a function of stimulus type via off-line as opposed to
on-line influences. These off-line influences of stimulus type involve differences in perceptual learning driven by differential
exposure to the different types of stimuli.
We follow other authors in proposing excessive feature integration as one key mechanism in crowding (Huckauf &
Heller, 2002; Pelli et al., 2004, 2007). In doing so, we follow
Pelli and colleagues (Pelli et al., 2004, 2007) and Levi (2008)
as describing crowding as a bottleneck for object recognition
that occurs during the mapping of visual features onto object
representations. In this account, it is the large overlapping
structure of integration fields that causes crowding (see Figure
1). When a flanking stimulus falls within the integration field of
the target, features from that stimulus are incorrectly integrated
with the features of the target, hence interfering with target
identification. Here we briefly present an extension of this
general account of crowding that postulates some specific
mechanisms that might be driving the interference. Our model
of crowding is shown in Figure 8. This model combines knowledge from research on crowding, as reviewed by Levi (2008),
and our own research on character string processing and letter
perception that was the motivation behind the present study. It
can be thought of as a specific implementation of the general
two-stage model of crowding described by Levi (2008).
Overlapping integration fields are implemented within a generic
interactive-activation framework for object identification (McClelland
& Rumelhart, 1981), illustrated here for the specific case of letter
stimuli. There are two ways that flankers can inhibit target identification in this model: (1) between-level feature-object inhibition,
and (2) within-level object-object inhibition. Any inappropriate
feature that falls in the integration field of a given letter detector
will inhibit the activity of that letter detector via bottom-up
feature-object inhibition. This same feature can also contribute to
activation in a different letter detector, and thereby further inhibit
activity in the correct letter detector via within-level object-object
inhibition. Figure 8 shows two versions of our interactiveactivation model of crowding, one that includes bottom-up featureobject inhibition (left panel), and one without (right panel). The
version without bottom-up inhibition is proposed in the light of
recent evidence suggesting the absence of such a mechanism in
Figure 8. An interactive-activation model of crowding. A target (T) and
two flanking stimuli (F, B) fall within a large peripheral integration field.
Features extracted at the first layer (large grey letters) send activation to
letter detectors that compete for identification of a single letter at that
location (small black letters with lateral inhibitory connections indicated by
small filled circles). Strength of feature-letter excitation and inhibition is
depicted by full (stronger) versus dashed (weaker) lines. The left-hand
panel shows a version of the model that incorporates bottom-up featureobject inhibition, and the right-hand panel a version with no bottom-up
inhibition. The model on the right further illustrates the possibility that
features of the target can also be compatible with the flanker stimuli, and
vice versa.
foveal letter perception (Rey, Dufau, Massol, & Grainger, 2009).
However, it is not clear how a model without such bottom-up
inhibition could account for the stronger interference generated by
pseudo-letter flankers compared with letter flankers on target letter
identification, a finding reported by Huckauf, Heller, and Nazir
(1999).
The effect of flanker similarity can be explained in this framework as increased within-level (object-object) inhibition caused by
the target stimulus providing bottom-up support for the similar
flanker. This is illustrated in the right panel of Figure 8, where
connections from the target letter feed activation into the letter
detectors corresponding to the flanker stimuli. Furthermore, given
the evidence that more similar flankers generate stronger crowding, it is clear that the differential effects of crowding on letters and
symbols found in the present study cannot be interpreted in terms
of different levels of visual similarity between targets and flankers.
Our letters are arguably visually more similar to each other than
our symbol stimuli, so any effects of target-flanker similarity
would be acting against our hypothesis that symbols are subject to
greater levels of crowding than letters.
Why did the number of flankers (one or two) not have additive
effects with symbol targets in our study? This result fits our model
of crowding according to which part (if not all) of the interference
CROWDING AND STIMULUS TYPE
from flanker stimuli is generated by competing object representations (see Kalarickal & Marshall, 1999, for such a model of
receptive-field dynamics). The dynamics of lateral inhibition in an
interactive-activation model, in which all activated object representations are in competition, are such that a single strong competitor can be just as effective (in terms of target-directed inhibition) as several weaker competitors (see Grainger & Jacobs, 1996,
for an example of such nonlinearities in lateral inhibition for word
stimuli). This can therefore account for why a single flanker can be
just as effective as two flankers, when these fall entirely within the
integration field of the target.
Crowding and Reading Development
Why are letters less sensitive to crowding than symbols? The
answer to this question is very likely related to the extensive
experience that we have, as skilled readers, in processing strings of
letters compared with strings of symbols. Tydgat and Grainger
(2009) hypothesized that when children learn to read they develop
a specialized system that is custom built to handle the very specific
nature of written words (see McCandliss, Cohen, & Dehaene,
2003, for one version of this hypothesis). Most notably, this system
would develop in order to optimize processing in the extremely
crowded conditions that arise with printed words, compared with
other visual objects that do not typically occur in such a cluttered
environment. One way to achieve such optimization is to establish
a bank of letter detectors that enable parallel independent identification of letters in a string (Dehaene, Cohen, Sigman, &
Vinckier, 2005; Grainger et al., 2006; Grainger & van Heuven,
2003). Because these letter detectors are designed to identify a
given letter at a relatively precise location in space, they do not
require the relatively wide integration fields necessary to achieve
location invariance in object recognition. A reduction in the integration field size of these letter detectors could be the mechanism
responsible for the reduced effects of crowding found with letter
stimuli compared with symbols. This was the central hypothesis
tested and supported by the results of the present study, and
illustrated in Figure 1 in the introduction.
Some additional support for this hypothesis was provided by
Atkinson, Anker, Evans, Hall, and Pimm-Smith (1988), who found
greater crowding effects with letter targets and a combination of
horizontal and vertical flankers in 3- to 4-year-old children compared with 5- to 7-year-old children. More recent research examining the development of reading fluency provides further support
for this hypothesis. Kwon and Legge (2006) found that crowding
effects were stronger in children than adults. Furthermore, Kwon,
Legge, and Dubbels (2007) found a strong correlation between the
development of reading speed and the size of the visual span
(number of letters that can be identified without eye movements).
Because Pelli et al. (2007) have shown that crowding is the critical
factor determining the size of the visual span, one can again
conclude that reduced crowding in letter strings is a key factor at
play in the development of reading fluency.
Finally, there is research showing that crowding can be modified
by training. Huckauf and Nazir (2007) reported that training participants to identity targets surrounded by two flankers reduces the
effects of crowding, but mostly for the specific stimuli, eccentricities, and spacing used during training. Their results suggested that
more extensive training might be necessary to observe greater
685
levels of transfer to untrained stimuli. This was done by Chung
(2007), who trained observers to identify target letters flanked by
two other letters with pretest and posttest phases separated by
6,000 training trials. The results of Chung’s study showed that
training produced a large improvement in letter identification
accuracy accompanied by an equally large reduction in critical
spacing (the inter-letter spacing allowing 50% correct target identification). However, given the huge exposure to letter strings of
the average adult participant in the above-cited studies, it seems
unlikely that even extensive training as in Chung’s study, is having
the same influence as years of exposure to print. What is missing
at present is a developmental investigation of crowding effects
from pre-readers to skilled readers. We hypothesize that the integration fields of letter detectors decrease in size during the process
of learning to read. Therefore, a measure of critical spacing for
letter trigrams (e.g., Chung, 2007; Pelli et al., 2007) should reveal
a gradual decrease in the size of integration fields as a function of
reading age.
A possible neurophysiological correlate of such developmental
changes in orthographic processing is the N1 or N170 ERP component. Maurer et al. (2006) measured N1 activity in kindergarten
pre-readers and the same children after 2 years of reading instruction (grade 2). The children were presented with words,
pseudowords, and strings of symbols. They found a significant
increase in N1 amplitude to word stimuli as a function of reading
experience, and no such developmental change for symbol strings.
Furthermore, the difference in N1 amplitude for words and symbols correlated significantly with a measure of reading fluency in
second grade.
Finally, in future research it will be important to investigate
differences in crowding effects to letter and digit stimuli, as well
as other types of non-alphanumeric stimuli. Tydgat and Grainger
(2009) showed identical serial position curves with letter and digit
strings, both showing the classic W-shaped curve that was significantly different from the inverted U-shaped curve found for
strings of symbols. In the light of the present results, this would
suggest that digit stimuli benefit from the same reduction in
crowding as letter stimuli due to similar conditions of exposure to
these stimuli as strings of elements (i.e., multi-digit numbers).
Nevertheless, some developmental differences between letters and
digits might be expected given the different amounts of exposure
to these two types of stimuli, and differences in the average length
of words and numbers.
Asymmetrical Crowding and the Initial
Position Advantage
In their analysis of differences in the serial position functions for
letters and symbols, Tydgat and Grainger (2009) observed a systematic advantage for the initial position of strings of letters
compared with symbols. While the advantage for the final position
in letter strings was found to depend on masking conditions, the
advantage for the initial position was robust across all conditions
for letters, and was never present for symbols. This suggests that
the initial position of strings of letters benefits from an additional
advantage over and above that of reduced crowding via smaller
integration field size. In an attempt to provide an integrated account of the complete pattern of effects found in their study,
Tydgat and Grainger proposed that not only does the size of
686
GRAINGER, TYDGAT, AND ISSELÉ
integration fields of location-specific letter detectors change with
exposure to printed stimuli, but that there is also a change in the
shape of these integration fields.
According to Tydgat and Grainger (2009), the hypothesized
change in shape of integration fields would arise in order to
optimize processing at the first position in strings of letters and
digits. Letters in the first position tend to provide more constraint
on lexical identity than any other position in the string (e.g., Clark
& O’Regan, 1999; Grainger & Jacobs, 1993). Furthermore, the
first position is critical for translating an orthographic code into a
phonological code, since the correct graphemes (letters and letter
clusters corresponding to a phoneme) can only be computed with
precise order information (Perry, Ziegler, & Zorzi, 2007). Such
optimization can be achieved by asymmetric integration fields that
are elongated in the direction of the initial position (i.e., to the left
for languages read from left-to-right). Letter detectors with integration fields of this shape will mostly suffer interference from
characters immediately to their left, hence endowing an advantage
for the initial position in strings. We further hypothesized that such
optimization only operates for stimuli that are fixated, and therefore only for detectors receiving input from the LVF.
This hypothesized visual-field specificity of the shape of integration fields for letter detectors accounts for why in Experiments
3 and 4 letter targets presented to the left of fixation showed
asymmetrical crowding effects with single flankers, whereas letter
targets presented to the right of fixation did not. Target letters
presented in the LVF showed stronger crowding from a single
flanker on their left, compared with the right. This is the equivalent
of an advantage for letters at the first position of a string of two
letters. This is exactly the pattern predicted by an elongation of
integration fields to the left for letter detectors in the LVF. Such
letter detectors suffer more interference from flankers falling to the
left of the target than flankers falling to the right. Furthermore, the
results of Experiment 5 showed that critical spacing values were
greater in the LVF than the RVF, and although the interaction
between visual field and target type was not significant, the difference between letters and symbols was numerically smaller in
the LVF. Because targets in Experiment 5 were always accompanied by two flankers (applying the standard critical spacing manipulation), this pattern is again in line with a hypothesized elongation of integration fields in the LVF.
Bouma (1973) was among the first to report visual field differences in letter identification in unpronounceable nonwords (note
that only the first and last positions were tested in Bouma’s study).
For a fixed eccentricity, there was a strong advantage for initial
over final position in the LVF which diminished and even reversed
(final position superior to initial position) in the RVF. This led
Bouma to suggest that “. . . the spatial extent of foveally oriented
masking is smaller in the R field than the Left” (p. 775). This is in
fact an alternative version of the present hypothesis. Indeed, there
is a general consensus in favor of an inward-outward anisotropy in
crowding effects for various types of stimuli (see Levi, 2008, for
review, and Petrov & Popple, 2007, for an example), and the
effects of single flankers with symbol targets in the partial report
task of Experiment 4 of the present study fits this general pattern.
This would suggest that in general (i.e., for different types of
stimuli) integration fields are elongated to the left in the LVF and
to the right in the RVF (Bouma, 1978). However, our results
suggest that the asymmetric effects found for symbol targets in
the RVF are driven primarily by location errors in the partial
report task, and this might explain the absence of such effects
in the 2AFC task (because the two alternatives were never a
flanking character). It might also be possible that stronger
asymmetries would arise with greater target eccentricities than
tested in the present study. The pattern found for letter targets,
on the other hand, did not depend on task, with a systematic
effect of flanker position arising in the LVF and not in the RVF.
This is in line with prior observations that the asymmetrical
effects of single flankers are stronger in the LVF for letter
stimuli (Krumhansl & Thomas, 1977; White, 1976). Current
research in our laboratory aims to plot the integration fields for
letter and symbol stimuli with independent manipulations of
left and right flankers. By separately measuring critical spacing
left and right of the target, we ought to be able to measure the
hypothesized asymmetric form of integration fields hypothesized in our model.
Conclusions
The present study investigated effects of crowding on the identification of letters and symbols using a standard flanker manipulation with eccentrically located targets. In line with the predictions of Tydgat and Grainger (2009), effects of crowding
interacted with target type, with stronger effects being found with
symbol targets compared with letter targets. When the number of
flankers was manipulated, this differential crowding effect was
found to be driven mostly by the effects of single flankers, which
always interfered with symbol identification but hardly affected
letter identification. Single flanker interference was found to be
robust for letter targets presented in the LVF when the flanker was
located to the left of the target. Finally, a manipulation of targetflanker spacing showed that symbols required a greater degree of
separation (larger critical spacing) than letters in order to reach a
criterion level of identification. These results are taken as evidence
in support of the hypothesis that during reading acquisition, the
size and shape of the integration fields of location-specific letter
detectors are modified, with an overall reduction in size with
increasing exposure to print, and a visual field specific modification of shape.
References
Atkinson, J., Anker, S., Evans, C., Hall, R., & Pimm-Smith, E. (1988).
Visual acuity testing of young children with the Cambridge Crowding
Cards at 3 and 6 m. Acta Opthalmologica, 66, 505–508.
Averbach, E., & Coriell, A. S. (1961). Short-term memory in vision. Bell
Telephone Technical Journal, 40, 19 –31.
Banks, W. P., Bachrach, K., & Larson, D. W. (l977). Asymmetry of lateral
interference in visual letter identification. Perception and Psychophysics, 22, 232–240.
Banks, W. P., Larson, D. W., & Prinzmetal, W. (l979). Asymmetry of
visual interference. Perception and Psychophysics, 25, 447– 456.
Bouma, H. (1970). Interaction effects in parafoveal letter recognition.
Nature, 226, 177–178.
Bouma, H. (1973). Visual interference in the parafoveal recognition of
initial and final letters of words. Vision Research, 13, 767–782.
Bouma, H. (1978). Visual search and reading-eye movements and functional visual field: A tutorial review. In J. Requin (Ed.), Attention and
performance, VII. Hillsdale, NJ: Lawrence Erlbaum Associates.
CROWDING AND STIMULUS TYPE
Butler, B. E. (1975). Selective attention and target search with brief visual
displays. Quarterly Journal of Experimental Psychology, 27, 467– 477.
Butler, B. E., & Merikle, P. M. (1973). Selective masking and processing
strategy. Quarterly Journal of Experimental Psychology, 25, 542–548.
Chastain, G. (1982). Confusability and interference between members of
parafoveal letter pairs. Perception and Psychophysics, 32, 576 –580.
Chung, S. T. L. (2007). Learning to identify crowded letters: Does it
improve reading speed? Vision Research, 47, 3150 –3159.
Clark, J. J., & O’Regan, J. K. (1999). Word ambiguity and the optimal
viewing position in reading. Vision Research, 39, 843– 857.
Dehaene, S., Cohen, L., Sigman, M., & Vinckier, F. (2005). The neural
code for written words: A proposal. Trends in Cognitive Sciences, 9,
335–341.
Ducrot, S., & Grainger, J. (2007). Deployment of spatial attention to words
in central and peripheral vision. Perception & Psychophysics, 69, 578 –
590.
Estes, W. K. (1972). Interactions of signal and background variables in
visual processing. Perception and Psychophysics, 12, 278 –286.
Estes, W. K., Allmeyer, D. H., & Reder, S. M. (1976). Serial position
functions for letter identification at brief and extended exposure durations. Perception and Psychophysics, 19, 1–15.
Grainger, J. (2008). Cracking the orthographic code: An introduction.
Language and Cognitive Processes, 23, 1–35.
Grainger, J., Granier, J. P., Farioli, F., Van Assche, E., & van Heuven,
W. J. (2006). Letter position information and printed word perception:
The relative-position priming constraint. Journal of Experimental Psychology: Human Perception and Performance, 32, 865– 884.
Grainger, J., & Jacobs, A. M. (1993). Masked partial-word priming in
visual word recognition: Effects of positional letter frequency. Journal
of Experimental Psychology: Human Perception and Performance, 19,
951–964.
Grainger, J., & Jacobs, A. M. (1996). Orthographic processing in visual
word recognition: A multiple read-out model. Psychological Review,
103, 518 –565.
Grainger, J., & van Heuven, W. (2003). Modeling letter position coding in
printed word perception. In P. Bonin (Ed.), The mental lexicon (pp.
1–24). New York: Nova Science Publishers.
Haber, R. N., & Standing, L. (1969). Location of errors with a poststimulus
indicator. Psychonomic Science, 17, 345–346.
Hammond, E. J., & Green, D. W. (1982). Detecting targets in letter and
non-letter arrays. Canadian Journal of Psychology, 36, 67– 82.
He, S., Cavanagh, P., & Intriligator, J. (1996). Attentional resolution and
the locus of visual awareness. Nature, 383, 334 –337.
Heron, W. (1957). Perception as a function of retinal locus and attention.
American Journal of Psychology, 70, 38 – 48.
Huckauf, A., & Heller, D. (2002). Spatial selection in peripheral letter
recognition: In search of boundary conditions. Acta Psychologica, 11,
101–123.
Huckauf, A., Heller, D., & Nazir, T. A. (1999). Lateral masking: Limitations of the feature interaction account. Perception and Psychophysics,
61, 177–189.
Huckauf, A., & Nazir, T. A. (2007). How odgcrnwi becomes crowding:
Stimulus-specific learning reduces crowding. Journal of Vision, 7, 18,
1–12.
Intriligator, J., & Cavanagh, P. (2001). The spatial resolution of visual
attention. Cognitive Psychology, 43, 171–216.
Kalarickal, G. J., & Marshall, J. A. (1999). Models of receptive-field
dynamics in visual cortex. Visual Neuroscience, 16, 1055–1081.
Krumhansl, C. L., & Thomas, E. A. C. (1976). Extracting identity and
location information from briefly presented letter arrays. Perception and
Psychophysics, 20, 243–258.
Krumhansl, C. L., & Thomas, E. A. C. (1977). Effect of level of confusability on reporting letters from briefly presented visual displays. Perception and Psychophysics, 21, 269 –279.
687
Ktori, M., & Pitchford, N. J. (2008). Effect of orthographic transparency on
letter position encoding: A comparison of Greek and English monoscriptal and biscriptal readers. Language and Cognitive Processes, 23, 258 –
281.
Kwon, M. Y., & Legge, G. E. (2006). Developmental changes in the size
of the visual span for reading: Effects of crowding [abstract]. Journal of
Vision, 6, 1003a.
Kwon, M. Y., Legge, G. E., & Dubbels, B. R. (2007). Developmental
changes in the visual span for reading. Vision Research, 47, 2889 –2890.
Levi, D. M. (2008). Crowding: An essential bottleneck for object recognition. A mini-review. Vision Research, 48, 635– 654.
Mason, M. (1975). Reading ability and letter search time: Effects of
orthographic structure defined by single-letter positional frequency.
Journal of Experimental Psychology, 104, 146 –166.
Mason, M. (1982). Recognition time for letters and nonletters: Effects of
serial position, array size, and processing order. Journal of Experimental
Psychology: Human Perception and Performance, 8, 724 –738.
Mason, M., & Katz, L. (1976). Visual processing of nonlinguistic strings:
Redundancy effects and reading ability. Journal of Experimental Psychology, 105, 338 –348.
Maurer, U., Brem, S., Kranz, F., Bucher, K., Benz, R., Halder, P., et al.
(2006). Coarse neural tuning for print peaks when children learn to read.
NeuroImage, 33, 749 –758.
McCandliss, B. D., Cohen, L., & Dehaene, S. (2003). The visual word form
area: Expertise for reading in the fusiform gyrus. Trends in Cognitive
Sciences, 13, 155–161.
McClelland, J. L., & Rumelhart, D. E. (1981). An interactive activation
model of context effects in letter perception: Part 1. An account of basic
findings. Psychological Review, 88, 375– 407.
Merikle, P. M., Coltheart, M., & Lowe, D. G. (1971). On the selective
effects of a patterned masking stimulus. Canadian Journal of Psychology, 25, 264 –279.
Merikle, P. M., Lowe, D. G., & Coltheart, M. (1971). Familiarity and
method of report as determinants of tachistoscopic performance. Canadian Journal of Psychology, 25, 167–174.
Mewhort, D. J. K., & Campbell, A. J. (1978). Processing spatial information and the selective-masking effect. Perception and Psychophysics, 24,
93–101.
Põder, E. (2008). Crowding and visual complexity. Perception (Supplement), 37, 121.
Pelli, D. G., Burns, C. W., Farell, B., & Moore-Page, D. C. (2006). Feature
detection and letter identification. Vision Research, 46, 4646 – 4674.
Pelli, D. G., Palomares, M., & Majaj, N. J. (2004). Crowding is unlike
ordinary masking: Distinguishing feature integration from detection.
Journal of Vision, 4, 1136 –1169.
Pelli, D. G., & Tillman, K. A. (2008). The uncrowded window of object
recognition. Nature Neuroscience, 11, 1129 –1135.
Pelli, D. G., Tillman, K. A., Freeman, J., Su, M., Berger, T. D., & Majaj,
N. J. (2007). Crowding and eccentricity determine reading rate. Journal
of Vision, 7, 1–36.
Perry, C., Ziegler, J. C., & Zorzi, M. (2007). Nested incremental modeling
in the development of computational theories: The CDP⫹ model of
reading aloud. Psychological Review, 114, 273–315.
Petrov, Y., & Popple, A. V. (2007). Crowding is directed to the fovea and
preserves only feature contrast. Journal of Vision, 7, 1–9.
Pitchford, N. J., Ledgeway, T., & Masterson, J. (2008). Effect of orthographic processes in letter position encoding. Journal of Research in
Reading: Special Issue –Orthographic Processes in Reading, 31, 97–
116.
Pitchford, N. J., Ledgeway, T., & Masterson, J. (2009). Reduced orthographic learning in dyslexic adult readers: Evidence from patterns of
letter search. The Quarterly Journal of Experimental Psychology, 62,
99 –113.
Rey, A., Dufau, S., Massol, S., & Grainger, J. (2009). Testing computa-
688
GRAINGER, TYDGAT, AND ISSELÉ
tional models of letter perception with item-level event-related potentials. Cognitive Neuropsychology, 26, 7–22.
Schwantes, F. M. (1978). Stimulus position functions in tachistoscopic
identification tasks: Scanning, rehearsal, and order of report. Perception
and Psychophysics, 23, 219 –226.
Stevens, M., & Grainger, J. (2003). Letter visibility and the viewing
position effect in visual word recognition. Perception and Psychophysics, 65, 133–151.
Tydgat, I., & Grainger, J. (2009). Serial position effects in the identification of letters, digits, and symbols. Journal of Experimental Psychology:
Human Perception and Performance, 35, 480 – 498.
Van der Heijden, A. H. C. (1992). Selective attention in vision. London:
Routledge.
White, M. J. (1976). Order of processing in visual perception. Canadian
Journal of Psychology, 30, 140 –156.
Wolford, G., & Hollingsworth, S. (1974). Retinal location and string
position as important variables in visual information processing. Perception and Psychophysics, 16, 437– 442.
Zhang, J. Y., Zhang, T., Xue, F., Liu, L., & Yu, C. (2007). Legibility
variations of Chinese characters and implications for visual acuity measurement in Chinese reading population. Investigative Ophthalmology
and Visual Science, 48, 2383–2390.
Received May 28, 2008
Revision received January 29, 2009
Accepted April 13, 2009 䡲
E-Mail Notification of Your Latest Issue Online!
Would you like to know when the next issue of your favorite APA journal will be available
online? This service is now available to you. Sign up at http://notify.apa.org/ and you will be
notified by e-mail when issues of interest to you become available!